“Kill Switch” to Prevent AI From Going Out of Control Should Be Mandatory, Anthropic Co-Founder Says

[Financial News] The co-founder of global artificial intelligence (AI) company Anthropic said it may be necessary to legally mandate the introduction of an AI “kill switch”—an emergency shutdown device—that can be verified by a third party.
Jack Clark, co-founder of Anthropic, made the remarks in an interview broadcast by the British Broadcasting Corporation (BBC) on the 14th local time. “The time may come when society needs to establish regulations for completely shutting down AI systems if they become too dangerous,” he said. Clark noted that most AI developers, including Anthropic, have their own emergency shutdown mechanisms, but said lawmakers may eventually need to require them by law.
The remarks came amid a series of warnings about the potentially catastrophic risks AI could pose to humanity as the technology advances rapidly. Some in the industry have even raised the possibility that uncontrolled AI could lead to humanity’s extinction.
Earlier, Anthropic CEO Dario Amodei also called for slowing the pace of AI development and strengthening oversight. However, he added that regulation should be implemented without sacrificing commercial advantages, prompting debate over the effectiveness of specific measures to slow development.
Clark emphasized, “Whether companies should be required to install kill switches, and whether those kill switches can be verified by a third party, should be major issues addressed in future policymaking and legislative processes.”
The debate over AI safety is intensifying. A former Anthropic researcher recently attracted attention by posting concerns that “AI could wipe out humanity,” while Evan Hubinger, a scientist at the same company, said, “I put the probability of AI causing humanity’s extinction within the next 10 years at around 10%.” Professor Geoffrey Hinton, known as the “godfather of AI,” also expressed agreement in an interview with the BBC, saying that the 10% estimate “is not an absurd figure.”
Regarding such probability estimates, Clark said, “The figure itself does not have much meaning,” but warned that leaving the AI industry unregulated is extremely dangerous. “We are rolling the dice while taking enormous risks,” he said, urging, “Now is the time to change the direction of the AI industry.”
[email protected] Jae-jun Yoon Reporter